inlierratio

Inlierratio, often referred to as the inlier ratio, is a measure used in robust estimation to describe the proportion of data points that are consistent with a given model relative to the total data. It is typically defined as w = N_inliers / N_total, where inliers are observations whose residual error with respect to the model falls below a predefined threshold.

In computer vision and related fields, inlierratio is a key parameter in algorithms such as RANSAC. It informs the expected likelihood of sampling inliers-only subsets and influences the number of iterations required to achieve a desired confidence level. For a sample size s, the required iterations can be expressed as k = ceil(log(1 - p) / log(1 - w^s)), where p is the target probability. The estimate of w may be updated as the algorithm progresses, based on current inlier counts.

Interpretation and limitations: A higher inlierratio suggests that a larger portion of the data conforms to the model, indicating a better model fit or less noise. However, the value depends on the chosen residual threshold and the assumed model; mis-specification or biased thresholds can distort w. In practice, inlierratio is used to evaluate model quality, to adjust termination criteria, and to guide sampling strategies in robust estimation workflows. It is related to the broader concept of data reliability and can interact with other measures such as outlier ratios and residual statistics.

Related concepts include outlier ratio, residual thresholding, and various robust estimators. The term appears in tasks such as 3D reconstruction, motion estimation, and feature matching, where robust model fitting is essential.